92ea9ee6f5
Establishes the canonical Delta-CRDT reference contract so a future native (Rust/Zig) backend is gate-G1-eligible under ADR-0196 — the ZC-0 'contract pinning' slice. No Zig code; ZC-1+ remains gated at G2. - vault/crdt.py: canonical Python reference (ArenaEntry, Delta, LocalArena, merge_kernel, canonical_bytes, delta_hash). Pure content law — content- addressed by IEEE-754 bits then provenance; no normalization, no versor closure, no global Vault writes. - ZC-0 contract tests (semilattice C-1..C-5; content ordering / signed-zero / NaN bit-addressing; C-7 no-global-write) — all failable (mutation-checked: no-dedup breaks C3/C5, arrival-order breaks C1). - Golden fixture corpus (tests/fixtures/crdt/) regenerated deterministically from the reference; single source of truth also emits the Rust expected hex. - core-rs: Delta::canonical_bytes + test_crdt_hash_parity.rs proving Rust produces byte-identical canonical_bytes to the Python reference. - ADR-0180 -> Accepted: locked contract, byte layout, obligation map, and the explicit boundary that no Zig is authorized. Verification: ZC-0 21 passed, Rust arena+parity 16 passed, architectural invariants 40 passed, smoke 67 passed. Serving frozen: 7/8 lane SHAs match; the public_demo miss is a pre-existing wall-clock budget overrun (ADR-0099, ~46-48s > 30s) reproduced identically on clean main — environmental.
47 KiB
47 KiB
Architecture Decision Records
This directory contains the Architecture Decision Records (ADRs) for the CORE project.
ADRs record significant architectural decisions: what was decided, why, what alternatives were considered, and what consequences follow. They are permanent records — superseded ADRs are archived, not deleted.
Index
| ADR | Title | Status |
|---|---|---|
| ADR-0001 | VocabManifold Versor Invariant | Accepted |
| ADR-0002 | Ingest Layer Architecture | Accepted |
| ADR-0003 | Coordinate System Dissolution | Accepted |
| ADR-0004 | Rotor as Operator, Not Vocabulary Property | Accepted |
| ADR-0005 | Language Pack Contract | Accepted |
| ADR-0006 | The Field Energy Operator (Hamiltonian Companion Field) | Implemented |
| ADR-0007 | The Valence Layer | Accepted |
| ADR-0008 | Allocation Physics | Accepted |
| ADR-0009 | Compositional Physics | Accepted |
| ADR-0010 | Identity Physics | Accepted |
| ADR-0011 | Renderer Layer Contract | Accepted |
| ADR-0012 | core_ingest Governance Layer |
Accepted |
| ADR-0013 | sensorium/ Multimodal Protocol Layer |
Accepted |
| ADR-0014 | train/ Learning Loop |
Accepted (Stub) |
| ADR-0015 | Language Packs as Compiled Linguistic Manifolds | Accepted |
| ADR-0016 | Capability Roadmap and Eval Methodology | Accepted |
| ADR-0017 | Agency Scope: Responsive-with-Axiology | Accepted |
| ADR-0018 | Tool Use Scope: Typed Deterministic Operators | Accepted |
| ADR-0019 | Exact Vault Recall Acceleration | Accepted |
| ADR-0020 | Phase 5 / Rust Parity Sequencing | Accepted (2026-05-16) |
| ADR-0021 | Epistemic Grade Policy | Accepted |
| ADR-0022 | Forward Semantic Control | Accepted (2026-05-17 — all five TBDs addressed; all |
| ADR-0023 | Forward Semantic Control: Proof Evidence | — |
| ADR-0024 | Inner-Loop Per-Rotor Admissibility | — |
| ADR-0025 | Rotor / Frame Admissibility | — |
| ADR-0026 | Ranked Admissibility with Margin | Accepted (2026-05-17) |
| ADR-0027 | Identity Packs — Load-Bearing, Swappable, Ratified | Accepted (2026-05-17) — Phases 1–6 complete; Phase 7 (this doc + the operational reference) complete; deep realizer wiring landed under ADR-0028 (2026-05-17). |
| ADR-0028 | Identity Surface Wiring — Pack-Driven Hedge & Claim Strength | Accepted (2026-05-17) |
| ADR-0029 | Safety Packs — Always-Loaded, Never-Replaceable Boundaries | Accepted (2026-05-17) |
| ADR-0030 | Depth-Language Hedge Wiring | Accepted (2026-05-17) |
| ADR-0031 | Score-Decomposition Surface — Per-Axis Hedge Phrases | Accepted (2026-05-17) |
| ADR-0032 | SafetyCheck — Structural Surface for Safety-Pack Boundaries | Accepted (2026-05-17) |
| ADR-0033 | Ethics Packs — Swappable Domain Commitments | Accepted (2026-05-17) |
| ADR-0034 | EthicsCheck — Structural Surface for Ethics-Pack Commitments | Accepted (2026-05-17) |
| ADR-0035 | Turn-Loop Verdict Surfacing for SafetyCheck and EthicsCheck | Accepted (2026-05-17) |
| ADR-0036 | Safety-Only Typed Refusal Policy | Accepted (2026-05-17) |
| ADR-0037 | Per-Predicate Ethics Refusal Opt-In | Accepted (2026-05-17) |
| ADR-0038 | Hedge Injection as a Runtime-Level Affordance | Accepted (2026-05-17) |
| ADR-0039 | Audit Completeness — TurnVerdicts Bundle, Stub-Path TurnEvent, hedge_injected Signal |
Accepted (2026-05-17) |
| ADR-0040 | Structured-Logging Sink for Turn-Event Audit | Accepted (2026-05-17) |
| ADR-0041 | core chat --show-verdicts + Sink Fan-Out |
Accepted (2026-05-17) |
| ADR-0042 | Audit Tour Demo — core demo audit-tour |
Accepted (2026-05-17) |
| ADR-0043 | Phase-2 pack measurements: claims → numbers | Accepted (2026-05-17) |
| ADR-0044 | Medical / clinical ethics pack (worked-example domain pack) | Accepted (2026-05-17) |
| ADR-0045 | Long-context recall: CORE vs transformer baselines | Accepted (2026-05-17) |
| ADR-0046 | PropositionGraph as Forward Admissibility Constraint | Accepted |
| ADR-0047 | Wire the Forward Graph Constraint into the Chat Hot Path | Accepted |
| ADR-0048 | Pack-Grounded Surface for Cold-Start DEFINITION / RECALL | Accepted |
| ADR-0049 | Intent Classifier Head-Noun Subject Extraction | Accepted |
| ADR-0050 | Pack-Grounded Surface for Cold-Start COMPARISON | Accepted |
| ADR-0051 | Trust-Boundary Hardening Pass | Accepted |
| ADR-0052 | Teaching-Grounded Surface for Cold-Start CAUSE / VERIFICATION | Accepted |
| ADR-0053 | Cognition Lane Closure: Dev-Driven Corpus Expansion + CORRECTION Acknowledgement | Accepted |
| ADR-0054 | Vault Recall: Matrix-Cache Indexing + Batched API; Holdout Split Wired | Accepted |
| ADR-0055 | Inter-Session Memory: Reviewed Discovery Promotion | Phase A + Phase B Accepted; Phase C Implemented (ADR-0056/0057); Phases D–E substantially landed |
| ADR-0056 | Contemplation Loop: Question Decomposition + Polarity + Domain Typing (Phase C1) | Accepted (implemented at 4eecf73, 2026-05-18) |
| ADR-0057 | Teaching-Chain Proposal + Review + Replay-Equivalence Gate (Phase C2) | Accepted |
| ADR-0058 | forward_graph_constraint: Engaged but Inert on Today's Cognition Lane |
Accepted |
| ADR-0059 | Correction-Pass Telemetry Emission | Accepted |
| ADR-0060 | CORRECTION Acknowledgement Carries the Corrected-Topic Lemma | Accepted |
| ADR-0061 | PROCEDURE Intent Routes to Pack-Grounded Surface | Accepted |
| ADR-0062 | Composed Teaching-Grounded Surface (Chain-of-Chains) | Accepted |
| ADR-0063 | Cross-pack surface resolver | Accepted |
| ADR-0064 | Cross-pack teaching chains | Accepted |
| ADR-0065 | OOV gradient + relations v2 (Plan Phase 2) | Accepted |
| ADR-0066 | Turn-level composition (Plan Phase 3) | Accepted |
| ADR-0067 | Cross-pack teaching chains (Plan Phase 4) | Accepted |
| ADR-0068 | Register pack class (Plan Phase R1) | Accepted |
| ADR-0069 | Realizer register parameter (Plan Phase R2) | Accepted (amended 2026-05-19) |
| ADR-0070 | Second ratified register pack: terse_v1 (Plan Phase R3) |
Accepted |
| ADR-0071 | Seeded surface variation + discourse markers (Plan Phase R4) | Accepted |
| ADR-0072 | Register telemetry + operator surface (Plan Phase R5) | Accepted |
| ADR-0073 | Anchor lens: substrate-driven substantive variation | Accepted (umbrella ratified; sub-ADRs L1.1–L1.4 each ratify in isolation) |
| ADR-0073a | Anchor lens content phase (Plan Phase L1.1) | Accepted |
| ADR-0073b | Anchor lens class + loader (Plan Phase L1.2) | Accepted |
| ADR-0073c | First non-trivial lenses + composer wiring (Plan Phase L1.3) | Accepted |
| ADR-0073d | Anchor-lens telemetry, CLI, and tour demo (Plan Phase L1.4) | Accepted |
| ADR-0074 | Orthogonality tour: anchor-lens × register composition demo | Accepted |
| ADR-0075 | Realizer slot-type guard (C1: coherence floor) | Accepted |
| ADR-0076 | Confirmation-Tag Normalization (C2) | Accepted |
| ADR-0077 | Substantive register knobs + register-tour gate strengthening (R6) | Ratified |
| ADR-0078 | Composer/Graph atom equivalence telemetry | Ratified |
| ADR-0078 | ADR-0078 Phase 1 — Pre-Implementation Planning Note | — |
| ADR-0080 | Contemplation Loop: self-interrogation without self-ratification | Accepted |
| ADR-0083 | Transitive Chain Surface (Bounded Multi-Hop Composition) | Accepted |
| ADR-0084 | Definitional Layer for Lexicon Packs | Proposed |
| ADR-0085 | Gloss-Aware CAUSE Composer | Accepted |
| ADR-0087 | Rhetorical Style as Selection Axis (Pre-Work for Writing Curriculum) | Proposed |
| ADR-0091 | Domain Pack Contract v1 | Accepted |
| ADR-0092 | Reviewer Registry v1 | Accepted |
| ADR-0093 | Domain Pack Contract v1 Implementation | Accepted |
| ADR-0094 | Proposal Source Provenance | Accepted |
| ADR-0095 | Miner-Sourced Teaching Proposals | Accepted |
| ADR-0096 | Fabrication-Control Eval Lane | Accepted |
| ADR-0097 | Mathematics-Logic Reasoning-Capable Ratification | Accepted |
| ADR-0098 | Demo Composition Contract | Accepted |
| ADR-0099 | Public Showcase Demo | Accepted |
| ADR-0100 | Physics Reasoning-Capable Ratification | Accepted |
| ADR-0101 | Systems-Software Reasoning-Capable Ratification | Accepted |
| ADR-0102 | Hebrew-Greek Textual-Reasoning Reasoning-Capable Ratification | Accepted |
| ADR-0103 | Fluency Lane Attachment for ADR-0102 | Accepted |
| ADR-0104 | Curriculum-Sourced Teaching Proposals | Accepted |
| ADR-0105 | Sealed Holdout Encryption via age | Accepted (2026-05-22) |
| ADR-0106 | Expert-Demo Promotion Contract | Accepted |
| ADR-0107 | mathematics_logic Expert-Demo Promotion: Deferred |
Accepted (decision: defer promotion) |
| ADR-0108 | Proposed-ADR Sequencing Post-ADR-0105 | Accepted |
| ADR-0109 | Lane-Shape-Aware Thresholds (ADR-0106 Amendment) | Accepted |
| ADR-0110 | mathematics_logic Expert-Demo Promotion |
Accepted |
| ADR-0111 | physics Expert-Demo Promotion |
Accepted |
| ADR-0112 | Runnable Expert-Demo Showcase | Accepted |
| ADR-0113 | Rename expert-demo → audit-passed; Reserve expert for Future Capability Tier |
Accepted |
| ADR-0114 | Expert-Capability Roadmap: GSM8K-Math First | Proposed |
| ADR-0114a | Anti-Overfitting Proof Obligations for expert Promotion |
Accepted (documentation-only; no code change) |
| ADR-0114a.2 | OOD-Ratio Auditor (Obligation #2 wired for B3) | Accepted |
| ADR-0114a.5 | Reasoning-Isolation Perturbation Suite (Obligation #5, B3) | Accepted |
| ADR-0114a.6 | Compositional-Depth Curve Auditor (Obligation #6 wired for B3) | Accepted (mechanism); coverage gap deferred to B3-owner follow-up |
| ADR-0114a.8 | Adversarial Generation Auditor (Obligation #8 wired) | Accepted (obligation passes; surfaces 2 known parser-layer gaps) |
| ADR-0114a.10 | Pack-Provenance Auditor (Obligation #10 wired for B3) | Accepted |
| ADR-0115 | Math Problem Parser and Typed Proposition Graph | Phase 1.1 Accepted (schema + 5 seed cases + tests); Phases 1.2–1.4 In Progress |
| ADR-0116 | Deterministic Solver (MathProblemGraph → SolutionTrace) |
Accepted |
| ADR-0117 | SolutionTrace Verifier |
Accepted |
| ADR-0118 | Stepped Realizer (SolutionTrace → Prose) |
Accepted |
| ADR-0118a | OOD Surface Generator for GSM8K-Style Parser Dev | Accepted |
| ADR-0119 | GSM8K Eval Lane Roadmap (Phase 5) | Proposed (roadmap-only) |
| ADR-0119.1 | Seal fabrication_control Holdout (ADR-0105 Amendment) | Accepted |
| ADR-0119.2 | GSM8K Eval Corpus Dev/Public Splits | Accepted |
| ADR-0119.3 | gsm8k_math Lane Runner (Phase 5.3) | Accepted |
| ADR-0119.4 | GSM8K Math: Frontier-Baseline Comparison (ADR-0114a §Obligation #7) | Accepted |
| ADR-0119.5 | Adversarial Generation (ADR-0114a Obligation #8) | Accepted |
| ADR-0119.6 | GSM8K Math Depth-Curve Measurement Harness | Accepted |
| ADR-0119.7 | Sealed GSM8K Test Set as gsm8k_math Holdout | Accepted |
| ADR-0119.8 | gsm8k_math Overall Lane Gate (gsm8k_capability_shape) |
Accepted |
| ADR-0120 | First expert Promotion Contract |
Proposed (contract-only; no domain promoted with this ADR) |
| ADR-0120 | ADR-0120 (math, ledger flip) — Mathematics-Logic Domain Promoted to expert |
Accepted — first expert-tier domain in the capability ledger |
| ADR-0120 | ADR-0120 (math) — Math-Expert Promotion Composer Wire-Up | Accepted (technical pass on first evaluation; awaiting reviewer signature for ledger admission) |
| ADR-0121 | mathematics_logic expert Promotion — Deferred (first attempt) |
Accepted (the deferral is the decision) |
| ADR-0122 | Parser Expansion: Rate / Per-Unit Reasoning (substrate-only; lift deferred) | Accepted (substrate landed; sealed-lift gate deferred — the |
| ADR-0122 | systems_software Audit-Passed Promotion: Deferred |
Accepted (decision: defer promotion) |
| ADR-0123 | Comparison-Phrasing Realizer (surface increment on the ADR-0123 substrate) | Accepted (surface increment; substrate landed in PR #155) |
| ADR-0123 | symbolic_logic Lane-Shape Remap (ADR-0109 Amendment) |
Accepted |
| ADR-0123a | all_three_pass_rate Synonym in inference_shape (ADR-0109 Amendment) |
Accepted |
| ADR-0124 | systems_software Audit-Passed Promotion |
Accepted |
| ADR-0125 | Reasoning-Isolation Perturbation Suite | Accepted |
| ADR-0126 | Candidate-Graph Parser with Round-Trip Verifier-Filter | Proposed |
| ADR-0127 | ADR-0127 + ADR-0128 Results — Path-B Triggered | Empirical result; load-bearing for the GSM8K-math arc decision |
| ADR-0127 | en_units_v1 Pack + Units-Aware Candidate Extractors |
Proposed (scope-only; implementation follow-up to ADR-0126) |
| ADR-0128 | en_numerics_v1 Pack |
Proposed (scope-only; sibling to ADR-0127) |
| ADR-0129 | Spaced Reviewed-Correction Replay (Deferred Proposal) | Proposed — Deferred (backlog item; no implementation |
| ADR-0130 | Pre-Articulation Calibration Logging (Deferred Proposal) | Proposed — Deferred (backlog item; no implementation |
| ADR-0131 | Re-Target Math Expert Promotion to Architecture-Aligned Benchmarks | Proposed |
| ADR-0131.1.F | B1 Symbolic Equivalence: Frontier-Baseline Comparison | Proposed |
| ADR-0131.1.S | Sealed Holdout for Benchmark 1 (Symbolic Equivalence v1) | Accepted |
| ADR-0131.2 | Benchmark 2: CORE-native teaching-corpus eval (lane gate) | Accepted |
| ADR-0131.2.B | Benchmark 2: B2 teaching-corpus enrichment (load-bearing gate) | Accepted |
| ADR-0131.3 | Benchmark 3: Bounded-Grammar Word Problems | — |
| ADR-0131.4 | Composite Math-Expert Promotion Gate (wired) | Accepted |
| ADR-0131.5 | GSM8K Coverage Probe: Retirement After G.x Axis Completion | Accepted |
| ADR-0131.G | GSM8K Coverage Probe: Honest Measurement Under the Safety Rail | Proposed |
| ADR-0131.G.0 | Probe Substrate: Candidate-Graph Pipeline | Proposed |
| ADR-0131.G.1 | Capability axis: state-introducing verb classes | Accepted |
| ADR-0131.G.2 | Capability axis: comparative operations (additive + multiplicative) | Proposed |
| ADR-0131.G.3 | Numeric Literals (money + hyphenated cardinals) | Proposed |
| ADR-0131.G.3.1 | Numerics extensions (fractions + multi-currency + multi-token cardinals + word-num-adjective) | Proposed |
| ADR-0131.G.4 | Capability axis: multi-clause composition (conjoined subjects, conjoined objects, embedded quantifiers) | Proposed |
| ADR-0131.G.5 | Aggregate Answer Composition | Accepted |
| ADR-0132 | Semantic-Symbolic Binding Graph: Phase 1 data model | Accepted (Phase 1 only; Phases 2–5 deferred) |
| ADR-0133 | Semantic-Symbolic Binding Graph: Phase 2 adapter from MathProblemGraph |
Accepted (Phase 2 only; Phases 3–5 deferred) |
| ADR-0134 | Binding Graph Phase 3: Unit-Aware Equation Admissibility | accepted |
| ADR-0135 | Binding Graph Phase 4: question-target binding refinement | Accepted. |
| ADR-0136 | Statement-Layer Corridor: Graduated GSM8K Admission via Parser Extension | Active — Regex sentence-template prescription superseded by ADR-0164 (2026-05-26). Empirical taxonomies preserved. |
| ADR-0136.S.1 | Rate/Event Statement Parsing | Accepted — regex patterns scheduled for removal under ADR-0164 Phase 3; closed-set vocabulary preserved as lexicon seed |
| ADR-0136.S.2 | Conditional-Op Question (Statement-Layer Corridor) | Active — regex patterns scheduled for removal under ADR-0164 Phase 3; closed-set vocabulary preserved as lexicon seed |
| ADR-0136.S2 | post-rescan — Refusal Rescan v2: Barrier-Shift Ledger | Accepted |
| ADR-0136.S3 | Compound Initial-Mutation Extractor | Accepted — regex patterns scheduled for removal under ADR-0164 Phase 3; closed-set vocabulary preserved as lexicon seed |
| ADR-0136.S3 | post-rescan — Refusal Rescan v3: Barrier-Shift Ledger | Accepted |
| ADR-0136.S.4 | Novel Initial-Form Subject-Slot Widenings | Accepted — regex patterns scheduled for removal under ADR-0164 Phase 3; closed-set vocabulary preserved as lexicon seed |
| ADR-0138 | Comparative-Reference Layer | Draft (design-only) |
| ADR-0139 | Arithmetic-as-Versor Spike: add Only |
Draft |
| ADR-0140 | CORE Trace Protocol v0 | Proposed |
| ADR-0140 | subtract as Inverse Translator + Additive Group Closure |
Draft |
| ADR-0141 | multiply as Dilator (Positive Non-Zero Multipliers Only) |
Draft |
| ADR-0142 | Epistemic State Taxonomy — First-Class Vocabulary | Accepted (integration deferred pending ADR-0144) |
| ADR-0143 | Teaching-Derived Structural Recognition via Multi-Resolution Anti-Unification | Accepted |
| ADR-0144 | PropositionGraph — Epistemic Carrier and Recognition Integration Gate | Accepted |
| ADR-0145 | Energy-Modulated Vault Surface Readback | Accepted |
| ADR-0146 | L10 Shape B Hybrid Engine-State Persistence | Accepted |
| ADR-0148 | Wire VaultPromotionPolicy into turn boundary | Accepted |
| ADR-0149 | Integrate DerivedRecognizer into CognitiveTurnPipeline | Accepted |
| ADR-0150 | Autonomous Inter-Session Contemplation | Accepted |
| ADR-0151 | Auto-Proposal Pipeline at Load | Accepted |
| ADR-0152 | Learning-Arc Demo (core demo learning-arc) |
— |
| ADR-0153 | TurnEvent trace_hash back-stamp (W-020a) | accepted |
| ADR-0154 | DerivedRecognizer producer wiring (W-020b) | accepted |
| ADR-0155 | CI contemplation runner (W-021) | scoping |
| ADR-0156 | Atomic engine-state checkpoint writes (W-022 / L10b.1) | accepted |
| ADR-0157 | Revision-mismatch warning on engine-state load (W-023 / L10b.2) | accepted |
| ADR-0158 | reboot_event audit trail entry (W-024 / L10b.3) | accepted |
| ADR-0159 | Contemplation Quality Eval Lane (W-025) | Accepted |
| ADR-0160 | CORE Workbench v1: operator/auditor UI before public chat | proposed |
| ADR-0161 | HITL Async Queue (W-009, L11) | Proposed |
| ADR-0162 | Workbench Design System (v1) | Proposed |
| ADR-0163 | F2 — Confuser Corpus: a discrimination probe, not a coverage target | Proposed (spec only — no code). Follow-on to ADR-0163 §F (the Track-B |
| ADR-0163 | Path to GSM8K mastery: candidate-graph admissibility via the contemplation/HITL corridor | Proposed — Phases B–E prescription superseded by ADR-0164 (2026-05-26) |
| ADR-0164 | Incremental Comprehension Reader (replaces regex sentence-template parsing) | Partially implemented (Phase 1 + 2 shipped; eval delta pending lexicon expansion) |
| ADR-0164.1 | Lexical Primitive Set Scope (seed registry for en_core_math_v1) |
Proposed |
| ADR-0164.2 | Pronoun / Entity Resolution Policy | Proposed |
| ADR-0164.3 | Cross-Sentence Reading State | Proposed |
| ADR-0164.4 | Phase 2 Statement-Frame Reader | Proposed |
| ADR-0165 | Regex Scope Rule: Lexemes Only, Never Grammar | Proposed |
| ADR-0166 | Measurement-Capability Sequencing Discipline | Proposed |
| ADR-0167 | Audit-as-Teaching-Evidence (Math Reader → Contemplation) | Proposed (scoping ADR; no code in this PR) |
| ADR-0168 | FrameClaim Ratification Doctrine | Proposed (doctrine/scoping ADR; no runtime mutation in this PR) |
| ADR-0168.1 | MathFrameClaimProposal Adapter | Proposed (design bridge; no runtime FrameClaim admission in this PR) |
| ADR-0169 | CompositionClaim Ratification Doctrine | Proposed (doctrine/scoping ADR; no runtime mutation in this PR) |
| ADR-0169.1 | MathCompositionClaimProposal Adapter | Proposed (design bridge; no runtime CompositionClaim admission in this PR) |
| ADR-0170 | Recognizer Injector Contract Widening | Proposed (scoping ADR; no runtime change in this PR) |
| ADR-0172 | Math-Domain Corpus-Decomposition Mechanism (Learning-Arc Analog) | Proposed (scoping ADR; no runtime change in this PR) |
| ADR-0173 | Workbench Ratification Trust Boundary | Accepted (W0 of the workbench-UI wave; doctrine only — UI |
| ADR-0174 | Held-Hypothesis Comprehension with Lookback and In-Loop Contemplation | Proposed |
| ADR-0175 | Calibrated Attempt-and-Eliminate Learning: Two Regimes Under wrong=0 | Proposed |
| ADR-0176 | Multi-Step Grounded Composition with Question-Targeting | Proposed |
| ADR-0177 | Cue-Precision Learning: from practice eliminations to trusted cue→op patterns | Proposed |
| ADR-0178 | ADR-0178 GB-3b — referent-aware accumulation chaining (scope) | Proposed (scope only — no code). Sub-phase of |
| ADR-0178 | Compositional Structure: Comprehension-Guided Multi-Step Derivation (Gap B) | Proposed |
| ADR-0179 | Extraction Richness: feeding the comprehension composer real quantities | Proposed |
| ADR-0180 | Delta-CRDT Sharded Substrate for Multimodal Concurrency | Accepted (2026-05-31 — G1 reference-contract lock) |
| ADR-0181 | CORE-native Audio Compiler over the Delta-CRDT Substrate | Proposed |
| ADR-0182 | Cross-composer disagreement pooling: refuse distractor-quantity confusers without a reactive cue rule | Proposed (spec only — no code). Follow-on to |
| ADR-0183 | Lawful Audio→Lexeme Path (stub) | Proposed (stub — placeholder to record the fork; not yet a full design) |
Current frontier
Current as of 2026-05-29 (ADR-0183).
The development frontier has moved beyond the audit-passed foundations (ADR-0091–0125) into native multi-modal execution and autonomous calibration. Key active arcs include:
- Auditory Modality & CRDT Substrate (ADR-0180, ADR-0181, ADR-0183): Shifting from text-only to concurrent multi-modal pipelines. Introduces a lock-free Delta-CRDT shared vault substrate to eliminate thread contention under continuous vision/audio ingestion, and ships
audio_core_v1as a deterministic acoustic compiler lowerable to unit-versors. ADR-0183 stubs the serving-time ASR boundary, ensuring lexical extraction is either text-side or deterministically decoded underwrong=0. - Calibrated Attempt-and-Eliminate Learning (ADR-0175): Establishes a formal precision-only reliability ledger and ratio-gate using the one-sided Wilson lower bound (
conservative_floor). It separates the safe serving regime (wrong=0) from a sealed practice regime where failures are treated as elimination signals to prune the solver's search. - Incremental Comprehension & Verification (ADR-0164, ADR-0174): Transitioned sentence extraction from brittle regex sentence templates to a deterministic incremental reader based on shift-reduce over semantic categories. The reader is backed by held-hypothesis evaluation, lookback, and in-loop contemplation.
- Measurement & Evaluation Discipline (ADR-0166, ADR-0163-F2): Establishes the
capability-before-measurementrule to ensure eval lanes are only built when operators exist and can admit at least one case. F2 defines a confuser corpus discrimination probe rather than a mere coverage target.
Proposed-but-unimplemented ADRs
The following ADRs carry a status of Proposed in the Index:
Core Infrastructure & Metadata
- ADR-0084 — Definitional Layer for Lexicon Packs
- ADR-0087 — Rhetorical Style Axis
- ADR-0129 — Spaced Reviewed-Correction Replay (Deferred Proposal)
- ADR-0130 — Pre-Articulation Calibration Logging (Deferred Proposal)
- ADR-0140 — CORE Trace Protocol v0
Auditory Modality & Delta-CRDT Concurrency
- ADR-0180 — Delta-CRDT Sharded Substrate for Multimodal Concurrency
- ADR-0181 — CORE-native Audio Compiler over the Delta-CRDT Substrate
- ADR-0183 — Lawful Audio→Lexeme Path (stub)
Expert Promotion & Evaluation (Math/GSM8K)
- ADR-0114 — Expert-Capability Roadmap: GSM8K-Math First
- ADR-0119 — GSM8K Eval Lane Roadmap (Phase 5)
- ADR-0120 — First
expertPromotion Contract - ADR-0131 — Re-Target Math Expert Promotion to Architecture-Aligned Benchmarks
- ADR-0131.1.F — B1 Symbolic Equivalence: Frontier-Baseline Comparison
- ADR-0131.G — GSM8K Coverage Probe: Honest Measurement Under the Safety Rail
- ADR-0131.G.0 — Probe Substrate: Candidate-Graph Pipeline
- ADR-0131.G.2 — Capability axis: comparative operations (additive + multiplicative)
- ADR-0131.G.3 — Numeric Literals (money + hyphenated cardinals)
- ADR-0131.G.3.1 — Numerics extensions (fractions + multi-currency + multi-token cardinals + word-num-adjective)
- ADR-0131.G.4 — Capability axis: multi-clause composition (conjoined subjects, conjoined objects, embedded quantifiers)
- ADR-0163 — F2 — Confuser Corpus: a discrimination probe, not a coverage target
- ADR-0163 — Path to GSM8K mastery: candidate-graph admissibility via the contemplation/HITL corridor
Incremental Comprehension Reader & Skill Acquisition
- ADR-0126 — Candidate-Graph Parser with Round-Trip Verifier-Filter
- ADR-0127 —
en_units_v1Pack + Units-Aware Candidate Extractors - ADR-0128 —
en_numerics_v1Pack - ADR-0164.1 — Lexical Primitive Set Scope (seed registry for
en_core_math_v1) - ADR-0164.2 — Pronoun / Entity Resolution Policy
- ADR-0164.3 — Cross-Sentence Reading State
- ADR-0164.4 — Phase 2 Statement-Frame Reader
- ADR-0165 — Regex Scope Rule: Lexemes Only, Never Grammar
- ADR-0166 — Measurement-Capability Sequencing Discipline
- ADR-0167 — Audit-as-Teaching-Evidence (Math Reader → Contemplation)
- ADR-0168 — FrameClaim Ratification Doctrine
- ADR-0168.1 — MathFrameClaimProposal Adapter
- ADR-0169 — CompositionClaim Ratification Doctrine
- ADR-0169.1 — MathCompositionClaimProposal Adapter
- ADR-0170 — Recognizer Injector Contract Widening
- ADR-0172 — Math-Domain Corpus-Decomposition Mechanism (Learning-Arc Analog)
- ADR-0174 — Held-Hypothesis Comprehension with Lookback and In-Loop Contemplation
- ADR-0175 — Calibrated Attempt-and-Eliminate Learning: Two Regimes Under wrong=0
- ADR-0176 — Multi-Step Grounded Composition with Question-Targeting
- ADR-0177 — Cue-Precision Learning: from practice eliminations to trusted cue→op patterns
- ADR-0178 — ADR-0178 GB-3b — referent-aware accumulation chaining (scope)
- ADR-0178 — Compositional Structure: Comprehension-Guided Multi-Step Derivation (Gap B)
- ADR-0179 — Extraction Richness: feeding the comprehension composer real quantities
- ADR-0182 — Cross-composer disagreement pooling: refuse distractor-quantity confusers without a reactive cue rule
Workbench UI
- ADR-0160 — CORE Workbench v1: operator/auditor UI before public chat
- ADR-0161 — HITL Async Queue (W-009, L11)
- ADR-0162 — Workbench Design System (v1)
Historical — accurate as of ADR-0125; see the Index for the complete, current list of domain promotions.
ADR-0080 (Contemplation Loop, Phase 1), ADR-0110 (math audit-passed), ADR-0111 (physics audit-passed), and ADR-0124 (systems_software audit-passed) have all landed — mathematics_logic, physics, and systems_software are at audit_passed=true; the contemplation loop emits read-only SPECULATIVE findings from frontier_compare reports. The remaining ratified domain (hebrew_greek_textual_reasoning) needs its own promotion ADR.
ADR-0122 attempted systems_software promotion and deferred honestly on a lane-shape contract mismatch (symbolic_logic output shape vs registered checker), resolved by ADR-0123's shape remap and successfully promoted via ADR-0124.
Open candidate directions (no ADR yet)
Historical — accurate as of ADR-0125; see the Index for the complete, current list.
- Multi-reviewer holdout governance and threshold signing. ADR-0105 seals holdout payloads with a single recipient identity; multi-reviewer governance remains future work.
Accepted reasoning-capable domains
Historical — accurate as of ADR-0125; see the Index for the complete, current list of domain promotions.
Per ADR-0106, audit_passed is contract-gated, not threshold-only: a domain row may carry audit_passed=true only when a reviewer-signed audit_passed_claims entry exists whose evidence-bundle digest reproduces byte-for-byte. ADR-0107 attempted the first worked promotion and the contract refused; ADR-0109 amended the threshold rules; ADR-0110 then successfully promoted mathematics_logic (the first domain at audit_passed=true); ADR-0111 promoted physics second without further contract change, retiring the "math-only" objection. The other two ratified domains remain at reasoning-capable pending their own promotion ADRs.
| Domain | Ratification ADR | Pack(s) | Evidence summary |
|---|---|---|---|
mathematics_logic |
ADR-0097 + ADR-0110 | en_mathematics_logic_v1 |
All nine ADR-0091 predicates pass; ledger row is audit-passed (first such promotion, ADR-0110); all three attached lanes meet ADR-0109 shape thresholds on public + holdout. |
physics |
ADR-0100 + ADR-0111 | en_physics_v1 |
All nine predicates pass; causal/modal operator coverage meets threshold; ledger row is audit-passed (second such promotion, ADR-0111); foundational_physics_ood 117/117 public + 39/39 holdout; shares inference_closure + fabrication_control results with math (distinct digest via domain_id). |
systems_software |
ADR-0101 + ADR-0124 | en_systems_software_v1 |
All nine predicates pass; transitive/causal operator coverage meets threshold; ledger row is audit-passed (third promotion, ADR-0124); all three attached lanes meet thresholds on public + holdout. |
hebrew_greek_textual_reasoning |
ADR-0102 + ADR-0103 | grc_logos_micro_v1, grc_logos_cognition_v1, he_logos_micro_v1, he_core_cognition_v1 |
First multi-pack ratification; all four packs carry uniform contract fields; causal/contradiction operator coverage meets threshold; ADR-0103 attaches Hebrew and Koine Greek fluency lanes with dev/public/holdout coverage. |
ADR chain notes
Forward Semantic Control closure — ADR-0022 through ADR-0026
ADR-0022 through ADR-0026 form a single coherent chain that closes forward semantic control as a non-stochastic, replay-deterministic, trace-evidenced mechanism.
- ADR-0022 establishes
AdmissibilityRegionand the contract that a region restricts the admissible token set at generation time. - ADR-0023 records proof evidence that the admissibility region is honored at the region-intersection level.
- ADR-0024 adds destination-side per-step admissibility and honest
InnerLoopExhaustionwhen all admissible candidates are rejected. - ADR-0025 adds rotor/frame admissibility when a region carries a
frame_versor. - ADR-0026 replaces static threshold tuning with ranked admissibility and a scale-invariant margin gate.
Runtime contracts for the chain are pinned in docs/runtime_contracts.md.
Pack-layer chain — ADR-0027 through ADR-0045
ADR-0027 through ADR-0045 establish the identity / safety / ethics pack architecture with deterministic remediation, audit completeness, telemetry, operator readout, audit-tour demo, pack measurements, a worked-example medical ethics pack, and long-context comparison measurements.
Evidence-governed domain chain — ADR-0091 through ADR-0111
ADR-0091 through ADR-0111 establish the current domain-ratification substrate and the audit-passed promotion gate that distinguishes contract-passing from demonstrated:
contract definition (0091)
↓
reviewer trust root (0092)
↓
validator / ledger enforcement (0093)
↓
negative-control fabrication lane (0096)
↓
reasoning-capable domain ratification (0097 / 0100 / 0101 / 0102)
↓
language-specific fluency lane attachment (0103)
↓
audit-passed promotion contract (0106 + 0109 amendment)
↓
worked audit-passed promotion (0110 — mathematics_logic, first)
↓
worked audit-passed promotion (0111 — physics, second; no contract change)
No domain claim should be treated as mature merely because a pack exists. reasoning-capable means the nine ADR-0091 predicates pass; audit-passed requires a reviewer-signed evidence-bundle digest that reproduces byte-for-byte from on-disk lane results.
The contract has been demonstrated end-to-end: refused once honestly (ADR-0107), amended once cleanly (ADR-0109), succeeded against mathematics_logic (ADR-0110), and succeeded against physics without further contract change (ADR-0111).
Comprehension-reader pivot — ADR-0164 and ADR-0165 (2026-05-26)
The GSM8K admissibility front-end is replaced. The regex sentence-template
parsing layer in generate/math_candidate_parser.py and
generate/recognizer_match.py is recognized as overfitting by
construction — it enumerates memorized surface shapes while pretending to
encode a closed grammar that English does not have.
-
ADR-0164 — Incremental Comprehension Reader. Word-by-word state accumulation over a closed set of semantic categories. The reader is a deterministic shift-reduce parser over categories, not over tokens. Output type is identical to the regex parser's output, so the binding-graph admissibility (ADR-0132/0133/0134/0135) downstream is unchanged. Operational lexicon lives in
language_packs/data/en_core_math_v1/alongside the existing packs. -
ADR-0165 — Regex Scope Rule. Structural invariant: regex is permitted only at the lexeme level (currency literal, fraction literal, percentage literal, time-amount, closed-set unit-noun), never at the sentence-structure level. The primitive set is a closed registry grown through the same contemplation → proposal → HITL review corridor that grows vocabulary.
These ADRs preserve every load-bearing piece of the prior work: the
binding graph (ADR-0132–0135), the solver / verifier / realizer
substrate (ADR-0116–0118), the capability-axis lanes G1–G5 and S1, the
HITL corridor (ADR-0150 / 0152 / 0155 / 0161), the wrong = 0 doctrine
and the replay-equivalence gate (ADR-0057, ADR-0114a). The
contemplation → proposal → review corridor architecture is reaffirmed
and its scope is generalized: it now ratifies lexicon entries,
categories, and lexeme primitives — not regex recognizers.
ADR-0163's diagnosis (the front-end is the bottleneck) is reaffirmed; its prescription (Phases B–E recognizer production) is partially superseded. ADR-0136 and its S-family have the same disposition.
Measurement-Capability Sequencing — ADR-0166 (2026-05-27)
Establishes the structural sequencing invariant: capability (operators, rules) must land on main and admit at least one test case before the eval lanes that measure it can be authored. Prevents noise of 100% refusing lanes from masking actual development priorities.
Calibrated learning and reliability — ADR-0175 through ADR-0177 (2026-05-28)
Establishes calibrated attempt-and-eliminate learning under a strict two-regime separation:
- Serving: Safe production path requiring
wrong=0via a strict, high-reliability Wilson lower bound gate. - Practice: Sealed evaluation path where mistakes serve as checkable elimination signals to prune solver paths.
Multimodal Audio modality & CRDT substrate — ADR-0180 through ADR-0183 (2026-05-29)
Shifts the engine toward concurrent multi-modal pipelines. Decouples physical lock contention on vision/audio from the logical manifold using Delta-CRDT sharding in Rust (ADR-0180), implements audio_core_v1 as a deterministic acoustic compiler (ADR-0181), and stubs ASR serving boundaries to prevent learned Whisper models from contaminating production serving paths (ADR-0183).
Session Logs
Session logs record the decisions and rationale from individual working sessions. They are not ADRs — they are the narrative record that informed the ADRs.
| Date | File |
|---|---|
| 2026-05-12 | SESSION-2026-05-12.md |
| 2026-05-12 (addendum) | SESSION-2026-05-12-b.md |
| 2026-05-12 (language packs) | SESSION-2026-05-12-language-packs-addendum.md |
| 2026-05-13 | SESSION-2026-05-13.md |
| 2026-05-26 (comprehension reader) | SESSION-2026-05-26-comprehension-reader.md |
| 2026-05-27 (parallel dispatch) | SESSION-2026-05-27-adr-0167-parallel-dispatch.md |
| 2026-05-27 (tier 3 sequencing) | SESSION-2026-05-27-tier3-sequencing.md |